On Jun 30, 2004, at 4:33 PM, Paul McNett wrote:
>> OK, postgres has a number of python access methods, it might
>> be worth splitting the access into a number of methods (we
>> use pgdb): just an idea.
>
> What are the pros/cons of each one? IOW, why is it that there
> are several dbapi libraries for postgres when one would
> suffice?
At PyCon DC 2004, there was pretty universal consensus that psycopg
was the adapter of choice. Things mentioned were dbapi-2.0 compliance,
speed, stability and support. Many of the adapters listed are no longer
actively developed/supported.
Here's an abstract from the PostgreSQL session that I went to at PyCon
DC 2004:
<http://www.sourcekeg.co.uk/www.python.org/pycon/dc2004/papers/23/>.
> I had originally set up the dBackend stuff to handle any number
> of backend libraries that could just be 'dropped in', but Ed
> didn't like that and switched to friendly names such as 'mysql'
> instead of 'MySQLdb' - IOW the name of the database instead of
> the name of the dbapi library.
The aim is to simplify. Developers should specify the backend, and
Dabo should handle the connection. We should default to a single
adapter per database. Those who are more adventurous can code their own
backends. I don't see any benefit in developing and supporting a dozen
Postgres adapters.
Since our volunteer is a pgdb user, then that's what we'll use
initially. Then if someone decides that some other adapter is better
and wants to develop the backend file and show us why, then we could
switch. But getting things to work in the first place is our main
objective.
___/
/
__/
/
____/
Ed Leafe
http://leafe.com/
http://dabodev.com/
©2004 Ed Leafe |
